Aaj Kay Lavayche: An Offline Garden Planner for Solapur, Powered by a Local Open Model

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

What I Built

Aaj Kay Lavayche ("what to plant today") is a small command-line garden planner for home gardeners in Solapur, Maharashtra. Solapur is hot and semi-arid, and most planting advice online is written for other climates, so I wanted something that fits where I actually live.

You run one command. It checks today's date, filters a list of 20 plants that suit Solapur (methi, kothimbir, palak, vangi, mirchi, tomato, bhendi, zendu, tulsi and more) down to what can be sown this month, and asks a local AI model for a short answer in simple Marathi: the top 3 plants for this week, a few sowing steps, and one watering tip for our climate.

The screen is meant to take about thirty seconds. Read the answer, put the phone or laptop down, and go plant something. A --container flag limits suggestions to pot-friendly plants for balcony gardeners, and a tiny garden log (--log "sowed methi in pot") lets me keep dated notes without opening any app.

Demo

[
]

Code

krisp619 / krisp619-Hacktoberfest-Open-Source-AI-Challenge-Week-01

# 🌿 Aaj Kay Lavayche (आज काय लावायचं?) > **An Offline Local-AI Garden Planner for Solapur, Maharashtra, India.** > *Built for Hacktoberfest 2026 Open-Source AI Challenge — Theme: "Touch Grass"* [](LICENSE) [](https://python.org) [](https://ollama.com) [](#) --- ## 🌻 What is "Aaj Kay Lavayche"? **"Aaj Kay Lavayche" (Marathi: *आज काय लावायचं?* / *"What should I sow today?"*)** is a lightweight, offline-first command-line garden planner tailored specifically for the hot semi-arid climate of **Solapur, Maharashtra**. Most gardening apps and AI assistants demand a persistent internet connection, trap you in endless infinite-scrolling feeds, or rely on expensive, latency-heavy cloud APIs. **Aaj Kay Lavayche changes that with a strict "Touch Grass" philosophy:** 1. Run **one simple command** from your terminal. 2. Get **three crisp recommendations**, sowing steps, and a Solapur-specific watering tip generated by a local open-weight LLM (or the built-in offline rules engine).…

View on GitHub

How I Built It

• Model: llama3.2:3b, an open-weight model, running locally through Ollama. It runs on an 8 GB laptop, and a --model flag lets me swap in any other local model.
• Data: a hand-written plants.json with sowing months, days to harvest, sunlight, water needs, whether the plant suits a pot, and a short tip for each plant. Anyone can edit it.
• Code: a small Python CLI with one dependency (the ollama package). It filters plants by month first and only then gives the short list to the model. That keeps the model's job small and its answer grounded in my own data, which is why a 3B model is enough.
• Fallback: if Ollama is not running or the model is missing, the tool prints a friendly error with the exact fix command and still shows the plain filtered list, so I never end up with nothing.
• Tests: pytest tests cover the month filter (including the December to January wrap), the container flag, the garden log and the error handling.

Why Does Open Innovation Matter?

• It works with no signal. A garden is not a place with reliable internet. After a one-time model download, the whole thing runs offline.
• My data stays with me. My garden notes and location never leave my laptop.
• It costs nothing to run. No API keys, no usage limits, no bills.
• I can change it. I can swap models, edit the prompt, change the answer language and correct the plant data for my own region. A closed API would not let me decide which data it trusts or how it behaves.